摘要
传统旋转机械结构故障诊断算法存在收敛慢、故障诊断精度低及迭代易陷入局部最优解的不足。本文在深度卷积神经网络训练故障数据的基础上,采用CS特征数据启发算法改善卷积神经网络的数据处理能力和泛化能力。CS启发算法引入适应度函数值自适应调整种群个体的移动步长,避免由于随机移动而导致全局范围内搜索精度的降低,提升故障诊断精度。实验结果表明,本文提出的CS启发算法具有更强的收敛性能和故障数据分类性能,对不同规模故障数据集诊断精度均优于三种现有机械故障诊断算法。
Traditional fault diagnosis algorithms for rotating machinery had some disadvantages,such as slow convergence,low accuracy of fault diagnosis and easy to fall into local optimal solution.Based on the fault data training based on deep convolution neural network,CS feature data heuristic algorithm was used to improve the data processing ability and generalization ability of convolutional neural network.CS heuristic algorithm introduces fitness function value to adaptively adjust the moving step size of population individuals,so as to avoid the reduction of global search accuracy caused by random movement and improve the fault diagnosis accuracy.The experimental results showed that the CS heuristic algorithm had stronger convergence performance and better fault data classification performance,and the diagnosis accuracy of different scale fault data sets was better than the three existing fault diagnosis algorithms.
作者
陈旭
CHEN Xu(Meishan Vocational and Technical College,Meishan 620010,China)
出处
《长春师范大学学报》
2023年第6期75-81,119,共8页
Journal of Changchun Normal University
